Effects of Keratin and Lung Surfactant Proteins on the Surface Activity of Meibomian Lipids

Abstract

Purpose.: In vitro studies indicate that surface tension and surface viscosity of the tear film lipid layer (TFLL) are governed by interactions between meibomian lipids and proteins from the aqueous layer. The role of minor tear proteins with strong lipophilic properties or those correlated with pathological states is still unknown. The discovery of lung surfactant proteins (SPs) in tears and keratin in normal and abnormal meibomian gland excretions warrants investigation into their effects on the surface activity of meibomian lipid films.

Methods.: Commercial keratin and bovine lung SPs were used in vitro to assess the surface pressure of meibomian lipid films using a Langmuir trough.

Results.: The pressure-area profiles of meibomian lipid films seeded with SPs (2.5 μL; 0.1 μg) demonstrated hybrid characteristics between meibomian lipid films alone and SPs alone but reached much higher maximum surface pressures (approximately 30 vs. 24 mN/m). Microscopically, the appearance of meibomian lipid films was not altered by SPs. Maximum surface pressure of meibomian films premixed with keratin was much higher than meibum alone. The pressure-area isocycles appeared more like those of meibomian lipids with a low concentration of protein and more like pure keratin films at a high concentration.

Conclusions.: The data strongly indicate that SPs and keratin likely interact with the TFLL. SPs are likely to act as strong surfactants and to reduce the surface tension of the lipid layer. Excess concentrations of keratin as identified in patients with meibomian gland dysfunction could disrupt the normal structure of the meibomian lipid film.

Introduction

Interference microscopy of the ocular surface indicates that the tear film lipid layer (TFLL) and its spread are important for the structural integrity of the whole tears. ^1,2 The TFLL is a self-assembled complex mixture of mainly nonpolar and some polar lipids derived from the meibomian glands. ^3–5 In addition to the lipids, it is believed that proteins form an integral part of the TFLL. ⁶ Since the aqueous component of the tear film has been shown to contain a vast array of proteins, ^7,8 some of these proteins probably adsorb to the TFLL from the aqueous component of the tear film. In vitro experiments indicate that such adsorption decreases surface tension. ^6,9,10 Other proteins are likely to be co-excreted with the lipids from the meibomian glands because approximately 90 different proteins ¹¹ have been identified in extracts expressed from meibomian glands. The general effect of proteins on meibomian lipids is to alter the structural order. Borchman et al. ¹² used principal component analysis to determine that bulk meibomian lipid samples from patients with meibomian gland dysfunction (MGD) contained higher levels of protein than normal meibum samples, and this resulted in a more viscous and ordered (less fluid) meibum. 

Interactions of lipid-binding proteins with the TFLL have been studied previously ^10,13,14 and reviewed in a special issue of IOVS. ¹⁵ Although surfactant proteins (SPs) as well as their possible interactions with the TFLL are extensively described in this review, ¹⁵ such interactions have not been investigated. The same can be said for very lipophilic proteins such as keratin. SPs are best known for their roles in stabilizing the lipids in lung surfactant, ^16–19 but they have also been found in extrapulmonary tissues ^20–22 including the lacrimal gland and tear film. ^23,24 In the lungs, there is considerable posttranslational modification of much larger peptides into the final smaller functional peptides, ^17–19 and so there may be differences between lung SPs and those found in tears. All four SPs (SP-A, -B, -C, and -D) have been detected in tears. ^23–25 Of these, SP-B and SP-C are of particular interest because these have the strongest surfactant property and hence the ability to interact with lipids. ^23,24 On the other hand, SP-A and SP-D, belonging to the C-type lectin family, are assigned by their complex immunological properties (for comprehensive review confer Wright et al. 2005 and van de Wetering et al. 2004). ^26,27 In the pulmonary system, SP-B and SP-C interact with phospholipids (mainly phosphatidylcholine) to form lung surfactant that lowers the surface tension at the surface of the terminal bronchioles and alveoli. ^18,19 During breathing, the lung SPs are adsorbed and desorbed into the lung surfactant and become incorporated into a multilayer arrangement, although the exact mechanisms involved are still unknown. ¹⁹ Given the mode of action of these SPs in the lung, it is reasonable to assume that they may have a similar role in the tear film. ^19,28 It is important to note that only a relatively small amount of surfactants is needed to affect the aqueous lipid interface; therefore, a relatively low amount in tears could still be of great functional significance. Another dimension to this idea has emerged from a recent publication indicating that Staphylococcus and Pseudemonas spp. are capable of synthesizing SPs, ²⁹ and hence this production of SPs during an ocular infection by these species might be a factor that could disrupt tear film stability. 

Keratins are the major proteins found in meibomian secretions ^11,30 and are very hydrophobic. They are formed by a cross-linked network of helical proteins and cystine-rich globular proteins. ³¹ The keratins in meibum most likely originate from the shedding of keratinized epithelial cells that line the ducts of the meibomian glands, ^30,32 but expression of keratin 7 and keratin 13 have also been found in meibomian acinar cells. ³³ Keratin may be important in the pathogenesis of obstructive MGD in which hyperkeratinization can cause up to a 10% increase in keratin compared with normal levels. ³⁰ This increase in keratin could add to the pathology of MGD by altering the structure of the TFLL. This idea has been supported in a recent study of meibum that suggested that the amount of protein in meibomian lipids increases the viscosity of meibum and contributes to the pathophysiology of MGD. ¹²  

Hence, in the current study, commercially available keratin and SP-B and SP-C extracted and purified from bovine lungs were used to assess the surface tension and pressure of meibomian lipid films in vitro. 

Materials and Methods

Chloroform, containing 1% ethanol as a stabilizer, was purchased from Sigma-Aldrich (Sydney, Australia); butanol, diisopropylether, dithiothreitol, keratin from human epidermis (batch 029K1252), methanol, polyacrylamide, sodium dodecyl sulfate, and other chemicals were all from Sigma-Aldrich. All chemicals used were of analytical grade. Water was purified using ion exchange and had a resistivity of 18 MΩ. 

Purification of Lung SPs

Lung SP-B and SP-C from cattle were used in these experiments because it is not possible to extract sufficient quantities from tears: both the volume of tears and the concentration of SPs in tears (∼2 μg/mL; Bräuer L, personal communication, 2012) are low. The lung SPs were purified using a method based on that of Beers et al. ³⁴ Permission from the local biosafety committee was given to obtain bovine lungs from a local slaughterhouse. Briefly, a bovine pluck was obtained, and the lungs were lavaged with approximately 10 L 50 mM sodium phosphate buffer (pH 7.4). The proteins were protected by processing the lavage at 4°C. The lavage was centrifuged at 600g for 10 minutes to remove large contaminants and cells, followed by a centrifugation of the supernatant at 19,500g for 35 minutes to precipitate the surfactant containing protein and lipids. One sixth of the precipitate (corresponding 1.6 L lavage) was resuspended in 10 mL purified water, the water phase extracted three times with 20 mL diisopropylether:butanol (3:2) to remove lipids, and the water phase containing lung SPs was freeze-dried. The freeze-dried sample was resuspended in 2 mL purified water and mixed with 40 mL chloroform–methanol (2:1) mixture. The lower organic phase was removed and stored, while the water phase and the interfacial fluff was re-extracted twice with 15 mL chloroform–methanol–water (86:14:1) according to Folch et al. ³⁵ The organic phases were pooled and the solvent removed in a rotary-evaporator. The dried precipitate was dissolved in 300 μL chloroform–methanol–HCl 1 N (1:1:0.008) and applied to a 30 × 1 cm Sephadex LH-20 (GE Healthcare, Rydalmere, Australia) column that had been equilibrated with the same solvent mixture. Fractions of 1.0 mL were collected at a flow rate of 0.5 mL/min and analyzed by UV/VIS spectrometry. All procedures were carried out using glassware to prevent contamination by organic extracts from plastics. 

Gel Electrophoresis of Bovine Surfactants

The size and purity of the SPs in each fraction from the column above were determined via sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) run under reducing and nonreducing conditions. Purified lung SP fractions (100 μL) were dried under nitrogen and reconstituted (20 μL of 0.35 M Tris/HCl pH6.8, 0.4% SDS). The samples were split and 5 μL of purified water was added to one part (nonreducing sample) and 5 μL 0.3 M dithiothreitol (reducing sample) to the other. Samples were heated at 94°C for 5 minutes before loading onto 15% polyacrylamide gels. Precision Plus Kaleidoscope standards (Bio-Rad, Sydney, Australia) were used as molecular weight markers. Following electrophoresis, gels were silver stained. ³⁶  

Western Blot Analysis of Purified Proteins for SPs

For Western blot analysis, 250 μL of the fractions was air-dried and resuspended in 60 μL SDS-PAGE sample buffer. Total protein was estimated by a Bradford assay using 1 μL of this solution. Six micrograms of total protein was then separated by 15% SDS-PAGE, transferred to nitrocellulose membranes (0.22-μm pore size) using a Towbin-buffer (13 mM Tris, 192 mM glycine) containing 20% methanol and 0.02% SDS. After transfer, the nitrocellulose membranes were blocked in TBST (66 mM Tris base, 137 mM NaCl, 2.7 mM KCl, and 0.1% Tween 20, pH 7.4) containing 5% milk powder for 1 hour at room temperature and then probed with 1:500 dilution of rabbit antibovine SP-B antibody (Abcam, Sapphire Bioscience, Waterloo, Australia) overnight at 4°C. After washing (three times) in TBST, the membranes were incubated with a 1:1000 dilution of goat antirabbit IgG-HRP (Cell Signalling Technology, Genesearch Pty. Ltd., Arundel, Queensland, Australia) for 2 hours at room temperature. The membranes were again washed in TBST and immunoreactive protein was visualized using Immobilon Western HRP substrate (Millipore, North Ryde, New South Wales, Australia). 

Preparation of Human Meibomian Lipids

Collection of meibomian lipids was in accordance with the tenets of the Declaration of Helsinki. Human meibomian lipids were gently squeezed out of the tarsal plates ¹⁰ of a single 56-year-old man with no external signs or symptoms of ocular pathology, including dry eye disease. Samples were pooled and stored in chloroform (1 mg/mL). 

Surface Pressure-Area (π-A) Measurements

Films were formed by spreading samples of purified SP preparations in elution buffer alone or premixed with human meibum at different ratios onto the cleaned surface (air–buffer interface) of an artificial tear buffer (NaCl: 6.63 g/L, KCl: 1.72 g/L, NaHCO₃: 1.38 g/L, CaCl₂·2H₂O: 0.15 g/L, NaH₂PO₄·H₂O: 0.10 g/L and MOPS: 4.18 g/L dissolved in ion exchange ultrapure water) ³⁷ in a double barrier 80-mL Langmuir trough (Nima Technology Ltd., Coventry, UK). The temperature of the subphase was controlled by a water jacket. Surface pressure-area (π-A) isocycles were recorded by measuring π using a Whilhelmy balance (Whatman filter paper No. 1; Whatman International, Maidstone, England), while cycling the surface area between 79 and 16 cm² at a rate of 10 cm²/min. Since we have previously found that films of meibomian lipids reorganize during isocycles and there are differences in the π-A curves at different temperatures, ³⁸ the following is a typical protocol. Lipids were initially spread at 20°C and isocycles carried out until the successive π-A profiles were the same (equilibrium isocycles). While continuing isocycles the subphase was heated to 35°C and when equilibrium isocycles were obtained at this temperature, the subphase was cooled back to 20°C and recordings continued until equilibrium isocycles were obtained. Experiments were repeated on a minimum of two different occasions for each mixture to verify. 

For experiments using keratin, keratin as supplied by Sigma-Aldrich (9.2 mg/mL in 8 M urea, 0.1 M mercaptoethanol, and 0.05 M Tris) was used. In this form, it is denatured and water soluble. In some experiments, it was applied directly to the surface between the barriers, either alone or premixed with meibomian lipids in a mass ratio of 0.4 to 1 up to 1.4 to 1 at its highest concentration. These are particularly high levels, given that the meibum would already have physiological levels of keratin, and were used to test what effects, if any, keratin might have on meibomian lipid films. It was found that the small amounts of keratin solution mixed well with meibomian lipids dissolved in chloroform. Since this form of keratin solution from Sigma-Aldrich was water soluble, the penetration of keratin from the subphase into spread meibomian lipid films was also determined. Human meibomian lipids (25 μL of 1 mg/mL) were spread onto the subphase between the barriers. and the film was aged by performing at least three isocycles. The barriers were then closed to a set surface area (same average area per molecule), which resulted in surface pressures of 5 mN/m (35°C) and 7 mN/m (20°C). Keratin was injected into the subphase just outside the barriers. The adsorption and penetration of keratin into the meibomian lipid layer was monitored over a 3- to 4-hour period. Isocycles were also conducted after the absorption of the protein to evaluate the change in surface activity of meibomian lipids after protein absorption or penetration, and fluorescent micrographs of the appearance of the films were also recorded. Assuming that keratin was evenly distributed in the 80-mL subphase, then the final concentration was calculated to be 0.115 μg/mL in the subphase for each microliter of keratin sample injected. The actual concentration of keratin in meibum has not been reported, and therefore the amounts of keratin used these experiments was a best guess based on a report of total protein in meibum of ∼0.24 μg per sample collected using a glass capillary tube (Thangavelu M, et al. IOVS 2010;51:ARVO E-Abstract 2373) and that keratin is <10% of the normal protein concentration. ³⁰ Assuming that at least 1 μL of sample was collected, then the concentration of total protein in meibum would be approximately 240 μg/mL. Our calculated final concentrations of keratin in the subphase were approximately 0.1, 7.4, and 10.6 μg/mL. Solutions of 8 M urea, 0.1 M mercaptoethanol, and 0.05 M Tris, equivalent to the solvents used to dissolve the Sigma-Aldrich keratin, were tested for surface activity as a control. 

Epifluorescence microscopy and an Andor Ixon high-speed camera (Andor Ixon model X-1323; Andor Technology plc., Belfast, United Kingdom) were used to examine the appearance of spread lipid films. For these experiments, the meibum was doped with 0.5% (w/w) of the fluorophore, 1-palmitoyl-2-[12-[(7-nitro-2-1,3-benzoxadiazol-4-yl)amino]dodecanoyl]-sn-glycero-3-phosphatidylcholine) (NBD-DPPC; Avanti Polar Lipids, Auspep Pty. Ltd., Tullamarine, Victoria, Australia). 

Results

Purification of Lung SPs

Optical density profiles of fractions obtained from the Sephadex LH20 column under nonreducing conditions had a peak centered on fraction 4 (Fig. 1), and SDS-PAGE of this fraction under nonreducing conditions resulted in bands at approximately 15 kDa and 25 kDa (Fig. 1A), which corresponded to the molecular weights of monomeric and dimeric forms of SP-B. ^23,39 A very faint third band, approximately 10 kDa, was also observed. This corresponded to the reported molecular weight of SP-C. ^39,40 An unidentified smeared band was observed in fraction 6 (lane 3, Fig. 1A). Under reducing conditions, two bands are visible at approximately 10 kDa (Fig. 1C), the monmeric SP-B, and at <10 kDa, which is the reported size of depalmitylated monomeric SP-C. This was supported by positive staining in Western blotting (Fig. 1D). Protein bands were not observed in fraction 2 (lane 1, Fig. 1A) or fractions 8 to 16 (lanes 4–8, Fig. 1A). On the basis of this evidence, it was determined that fraction 4 comprised SP-B and SP-C. The concentration of the purified bovine SPs was estimated to be approximately 0.39 μg/μL using a Bradford assay and bovine serum albumin as the standard. This fraction was used for surface activity experiments. 

Figure 1

 

(A) SDS 15% polyacrylamide gel of purified lung surfactant fractions separated under nonreducing conditions and stained with silver nitrate. Lane M, protein standard markers and the number above the lanes indicates the fraction number. (B) SDS 15% polyacrylamide gel of purified lung surfactant fractions separated under reducing conditions and stained with silver nitrate. Lane M, protein standard markers and profiles from fractions 2 and 4 are shown. Circled bands represent SPs. (C) Optical density profile of lung surfactant fractions separated using Sephadex LH20 column. (D) Western blot of lung surfactant (lane 1) and fraction 4 (lane 2).

Effect of Bovine Lung SPs on the Surface Activity of Meibomian Lipids

SP (10 μL; 4 μg, fraction 4) was applied between the barriers to the surface of the subphase in the trough at 20°C, and isocycles were carried out until equilibrium was reached. Isocycles were continued as the subphase was heated to 35°C, allowed to equilibrate, and then cooled back to 20°C. Except for the first isocycle, just after the film had been spread, the π-A profiles were almost identical (i.e., it reached equilibrium immediately). At 20°C, area at takeoff was less than that found in the first cycle. Heating to 35°C slightly increased hysteresis, and cooling back to 20°C returned the profile to that seen before heating. π_max was almost identical under all conditions (Fig. 2A). 

Figure 2

 

Comparative isocyles at different temperatures of 3.9 μg lung SPs alone (A) and of 0.1 μg lung SPs (B) or 0.2 μg SPs (C) each mixed with 20 μg meibomian lipids. Shown are the first (dashed line) and equilibrium isocyle (solid line) after spreading at 20°C, and equilibrium isocycles after heating the trough to 35°C (dotted line) and after cooling it back to 20°C (dashed/dotted line). The inserts show the scaled shapes of the individual π-area isocycles separated vertically for visual comparison.

The seeding of meibomian lipid films with small amounts of SPs at 20°C (2.5–5 μL; 1–2 μg) had a strong influence on the π-A profiles of meibomian lipid films and showed hybrid characteristics. With a smaller ratio of SPs, the π-A profiles were dominated by meibomian lipid characteristics (Fig. 2B) and differed from meibomian lipid profiles alone at 20°C ^10,38 in that π_max was much higher (∼30 vs. ∼24 mN/m^10,38) and there was not a steady increase in π_max with progressive isocycles (with pure meibomian lipids there is a gradual increase). The π-A profiles also showed the characteristics of meibomian lipid films such as reduced π_max on heating and increased surface pressures on cooling. With higher ratios of lung SPs (Fig. 2C), the hysteresis in the curves was markedly decreased, and similar to low ratios of lung SP, it reached a stable π_max quickly. There was still a slight decrease in π_max on heating to 35°C and a slight expansion on cooling, but far less than that seen with pure meibomian lipid films. 

Microscopically (Fig. 3), mixed films of SPs and meibomian lipids were more fluid (based on the movement of the film viewed under the microscope) at equivalent surface pressures than meibomian lipid films alone and overall seemed to be more even. However, for the first few isocycles, there were distinct dark patches at high pressures, which were not apparent once a number of isocycles had been completed. In Figure 3, the same feature can be seen in the first and second isocycles at around 30 mN/m. In the first isocycle, the dark patches seem to be mixed with the fluorescent lipids (hence gray in color) and are compressed as the pressure increases. In the second isocycle, the dark patches are further apart and more distinct. If the dark spots are local concentrations of SPs since these were not fluorescently labeled, an explanation for this could be that these molecules are unfolding at the surface. This would also indicate that some of the SPs and the meibomian lipids had not mixed completely, even though they were thoroughly mixed as a solution before being applied to the subphase. An alternate explanation is that they are local condensations of nonfluorescently tagged lipids, but this is less likely since such features in meibomian lipid films have never been observed. Upon heating to 35°C, the film was very fluid and its features could not be discerned, even at the highest surface pressure. When cooled back to 20°C, the film looked like a normal meibomian lipid film except that the equivalent appearances occurred at much higher surface pressures (compare with Fig. 6 pressure 21 in Millar et al. ¹⁰ with “Cooled 36” Fig. 3 here). 

Figure 3

 

Micrographs showing the appearance of a mixture of lung SPs (0.2 μg) and meibomian lipids (20 μg) seeded with a fluorescent marker. Isocycles were carried out at 20°C, then at 35°C then cooled back to 20°C. Note, for example, “26 iso1” means the micrograph was taken at a π of 26 mN/m during isocycle 1. The blurriness seen in “7.6 iso1 20°C,” and “24 35°C” indicate that the film was moving very fast (shutter speed 0.01 seconds). The arrows in “26 and 33 iso1 20°C” and in “30 iso2 20°C” indicate the same feature at different πs and isocycles. The arrows in “24 iso2” and “Cooled 34 iso5” highlight some subtle substructures in a relatively even film. Scale bar: 100 μm.

Surface Activity of Keratin and Keratin Mixed With Meibomian Lipids

Keratin (92 μg) spread onto the air–buffer interface at 20°C demonstrated strong surface activity with a π_max of approximately 30 mN/m (Fig. 4A), similar to that seen when initially applied at 35°C (not shown). The profile of the first isocycle began at a higher pressure and the initial compression led to a reorganization of the film to a lower pressure at maximum area. A notable hysteresis occurred and the films were very stable irrespective of changing the temperature. The solvent (control) showed no surface activity (not shown). When keratin (9.2, 28, or 37 μg) was premixed with 25 μg meibomian lipids and applied to the surface of the subphase, the π-A isocycles appeared more like meibomian lipids with the lower concentration of protein and more like pure keratin with the higher concentration (Figs. 4B–D). However, π_max was much higher than that achieved with meibomian lipids alone, and the films still showed expansion on cooling typical of meibomian lipid films. These observations indicate that the keratin and meibomian lipids were acting independently at the surface. The appearance of these films (Fig. 5) when first applied also suggests that this may be the case. Films containing 37 μg keratin, at low pressures and early isocycles, were stripy indicating lipid-rich areas (light) and lipid-poor areas (dark). We speculate that the keratin was in the dark areas. With further isocycles and at higher pressures, the film became patchy indicating that there had been clumping and disorganization of the lipids. This patchiness remained when the film was heated and cooled. By contrast, if only 9.2 μg of keratin was used in the mixture, then the film appeared much more like pure meibomian lipid films, with some dark spottiness once the film had been cooled to 20°C. With 28 μg of keratin, the mixed films looked more similar to the films with 37 μg of keratin (Fig. 5). 

Figure 4

 

Comparative isocycles at different temperatures of (A) keratin alone (92 μg) applied to the subphase surface; and (B–D) mixtures of different amounts of keratin (9.2 μg, 28 μg, 37 μg) with meibomian lipids (25 μg). The proteins and lipids were applied at 20°C, then heated to 35°C, and cooled back to 20°C. The inserts show the scaled shapes of the individual π-area isocycles separated vertically for visual comparison.

Figure 5

 

Micrographs showing the appearance of different amounts of keratin (37 μg, 9.2 μg) mixed with meibomian lipids (25 μg) seeded with a fluorescent marker. Isocycles were carried out at 20°C, then at 35°C, and then cooled back to 20°C. In the text boxes, the first number indicates π. The isocycle number temperature and amount of keratin are also shown. Scale bar: 100 μm.

Figure 6

 

(A) Penetration of different amounts of keratin into meibomian lipid films at different temperatures. The initial π was set at 5 to 8 mN/m and the keratin was injected into the subphase; 10 μL keratin resuspension buffer served as control. (B, C) Equilibrium isocycles at different temperatures after the penetration experiments with 64 μg keratin at 35°C and 92 μg keratin at 20°C, respectively.

Penetration of Keratin Into Meibomian Lipid Films and Its Effect on Surface Activity of the Meibomian Lipid Films

Before injection of keratin into the subphase, the meibomian lipid film was compressed to a target surface pressure and the surface area remained constant. An initial drop in surface pressure up to 1.5 mN/m was observed after the target pressure was reached due to relaxation of the meibomian lipid film. The rate and amount of penetration of keratin into the meibomian lipid films increased with concentration and temperature (Fig. 6A). This is typical for most proteins. ^6,10,41 Following penetration, isocycles were carried out (Figs. 6B, 6C). For the penetration at 35°C, after initial isocyles at that temperature, additional isocyles were carried out after the film had been cooled back to 20°C. Similarly, for penetrations done at 20°C, further isocycles were carried out at 35°C and then again after cooling back to 20°C. Irrespective of the temperature at which the penetration occurred, the curves of π-A isocycles at 35°C were almost identical, as were the curves following cooling back to 20°C (Figs. 6B, 6C). The shape of the curves more closely resembled the pure keratin curves (Fig. 4A) rather than the mixtures of meibum and keratin applied to the surface. In particular, they had a large hysteresis on expansion (Figs. 6B, 6C). Although this would suggest that keratin had displaced the meibomian lipids from the surface, the increase in maximum pressure upon cooling is a feature of meibomian lipids and not keratin. 

Micrographs taken prior to penetration at 20°C illustrate an amorphous mass and exhibited a few small, darker patches. Initial visual evidence of penetration was the appearance of some dark patches in the film followed by a progressive patchiness of the film until a very patchy film was observed at the end of penetration (Fig. 7). This indicated that there were concentrated areas of lipids (lighter regions) separated by darker regions where the keratin levels were presumably higher. After penetration, the films were noticeably stiffer and slower moving. During isocycles, the film retained the patchy appearance and particularly at high pressures appeared to have a stringy appearance (Fig. 7). This was very different from the stripy appearance of the mixed films (Fig. 5) but somewhat similar to the mixed films at high pressure after the films had been cooled. When heated to 35°C, the film was still patchy, but the lighter, (lipid) patches were dominant. At 35°C the films were moving too quickly to photograph even after penetration of keratin. After they were cooled, the films were quite different from those where the penetration had occurred at 20°C. The lighter lipid component appeared to be stringy and the darker possibly keratin component tended to occur in large patches. 

Figure 7

 

Micrographs showing the appearance of films of meibomian lipids (25 μg) seeded with a fluorescent marker before and after 92 μg of keratin was injected into the subphase (80 mL) and allowed to penetrate the spread lipid layer. The top row shows the appearance of a meibomian lipid film after 15 isocycles and set at 9 mN/m. As the keratin penetrated, it increased π to 18.1 mN/m and the film changed appearance (row 1). Isocycles were then carried out at 20°C (row 2), then at 35°C (row 3), and then after cooling back to 20°C (row 4). The bottom row shows the appearance of the meibomian lipid film after penetration of keratin at 35°C. Isocycles were carried out at 35°C and then after cooling back to 20°C. The figure boxes give π (first number), the temperature, and the number of isocycles. Scale bar: 100 μm.

Discussion

These data strongly indicate that SPs and keratin likely interact with the TFLL, but do so in very different ways. Supporting the idea for SPs was that the concentrations of SPs used in these experiments were not excessive. The amount of SPs used was similar to that in tears based on quantified concentrations of SP-B and SP-C in the tear film using ELISA (Bräuer L, unpublished data, 2012). However, in the experiments reported here, the possible effects of phospholipids in the aqueous proportion of the tears ⁴² was not considered. Given that SPs bind strongly to phospholipids, the actual concentration of free SPs available to bind to the lipid layer might be much less than the concentration that has been measured in whole tears. Nevertheless, if they were acting very strongly as surfactants, then it would be expected that only very small amounts would be needed to affect the meibomian lipid films. Hence, it would be worthwhile in future experiments to examine the effects of smaller amounts of SPs on meibomian lipid films. 

Currently, it is thought that as the alveoli in the lungs collapse and expand, the SPs are desorbed and adsorbed to the lung surfactant, ^16,19 and it is worthwhile considering if our data indicate that a similar process may be occurring in the TFLL. As background, it must be kept in mind that the lipid components of lung surfactant are polar and almost entirely phosphatidylcholine, ¹⁷ which contrasts to the composition of meibum, which is primarily waxes and cholesterol esters. ^4,5 The consequence of this fundamental difference is that phospholipids tend to form a monolayer on the surface and under high pressure collapse into a bilayer in the form of vesicles that move into the subphase, ^43,44 whereas meibum forms a duplex film with molecules that move off the surface going to lenses at the air interface. ⁴⁵ Our data suggest that the protein molecules are moving from the subphase into the lipid layer and at a minimum this can be interpreted as reducing the area per molecule occupied by the lipids. This puts pressure on the lipids to move off the surface. This is similar in principle to slowly adding lipids to the surface while maintaining a fixed area as was done by MacDonald and Simon ⁴³ using phosphatidylcholine. The transition from monolayer to bilayer begins for phospholipids at π ∼ 15 mN/m, and at approximately 50 mN/m all of the lipids exist as a bilayer. ^43,44 This leads to a curve showing the relationship between either the average molecular area per molecule or surface pressure and the ratio between the lipids existing at the surface and those desorbed from the surface as a bilayer (Fig. 3 in Feng et al. ⁴⁴ ). The monolayer and bilayer lipids are in equal amounts at approximately 29 mN/m. ^43,44 Once desorbed in this manner, the phospholipid vesicles do not return to the surface as a monolayer in the timeframe typical of isocycles used here. ^38,43 With films of lung surfactant, a similar desorption into the subphase is represented as a flattening of the curve during isotherms ¹⁶ and is again modeled as multilayer formation, which allows layers of phospholipids to be pushed into the subphase stabilized by SPs. These layers can return to the surface, but many are released into the subphase as unilamellar vesicles to be recycled in situ by macrophages and alveolar type II cells. Compared with lung surfactant, the SPs interact very differently with meibomian lipids. The constant and increased pressure during isocycles of mixed films of SPs and meibomian lipids indicate that they are not being desorbed and reabsorbed. If this had been the case, a flattening of the curve at smaller surface areas and a π_max similar to that of meibomian lipids alone would have been expected because this would represent desorption. With meibomian lipids, the stability of the isocycles through different temperatures and a similarity in appearance to normal meibomian lipid films suggest that they are acting as true surfactants in the TFLL and remain at the aqueous lipid interface. Alternatively, they may be desorbed into the upper phase of the lipid layer where they would most likely form lenses, but we are unable to determine if this was the case using our techniques. Nevertheless, the presence of SPs would lower the surface tension of the TFLL, even in low concentrations, and if the concentrations of SPs were increased, meibomian lipids appear to be able to tolerate this. However, this view might not reflect what is happening in vivo because of high compression rates and ratios during a blink. It appears that by rapidly compressing a film of bovine lung surfactant extract coating the surface of a captive bubble, the proteins stay in the film and the film does not collapse. Instead it becomes metastable (glass). ⁴⁶ Similar rapid compression experiments using meibomian lipids might resolve this. 

Similar arguments regarding possible desorption of lipids from the surface during adsorption of keratin could be applied and hence will not be repeated. In addition, the observations of mixed films of keratin and meibomian lipids indicate that excess keratin is likely to be disruptive to the normal structure of the lipid film given the change in the microscopic appearance of the film. This is evident by the eventual clumpy appearance of the lipid films when mixed with 37 μg of keratin. At the same time, the smoother looking lipid films when 9.2 μg of keratin was applied and the relatively normal appearance of the π-A isocycle profiles indicate that meibomian lipids are relatively tolerant of small amounts of keratin. It has been hypothesized that keratin may have a role in dry eye associated with MGD, but this is normally associated with its presumed role in obstructing secretion of meibum from the ducts due to hyperkeratinization. It has been shown that under these conditions the levels of keratin in meibum increase up to 10%. ³⁰ If, as indicated by the studies here, the keratin is mixed into the lipid layer of the tear film, it is likely that it would make the films more rigid and more subject to fracture under the blinking cycle. We have observed the appearance of the lipid layer of a long-term contact lens wearer with recently diagnosed contact lens–related MGD, and it was very similar in appearance to the speckled patterns presented in the micrographs in this study. A clinical study to evaluate the keratin composition in the lipid layer of contact lens wearers would be useful to determine the veracity of this observation. 

A note of caution is required with these studies in that we did not use human tear SP, but instead bovine lung SP. There is considerable posttranslational processing of lung SPs, including cleavage. ^23–25 Such posttranslational processing of tear SPs has yet to be determined, and therefore tear SPs may have slightly different characteristics ⁴⁷ from those shown here. Similarly, the keratin used in these studies was a denatured commercial product and in much higher than physiological concentrations. Indeed, keratin in its native form will not form surface films on an aqueous surface. ⁴⁸ This is perhaps highlighted by the fact that we were unable to find any previous literature that measured surface activity of keratin films on an aqueous surface. Also, although keratin is reported as a component of meibomian lipids, its conformation in the TFLL is unknown because its conformation when released onto the ocular surface is unknown. In terms of the mixed films with meibum, the meibum would have already contained a physiological level of keratin and so adding more keratin meant that these films had excess amounts. In vivo, it is possible that some extra keratin could come from the epithelial surface and adsorb to the TFLL, but it is unlikely that it would be at the concentrations used here. Nevertheless, the exaggerated concentrations used in these experiments indicated that keratin has a tendency to stiffen meibomian lipid films. In future experiments, it would be worthwhile to test the effects of lower concentrations of keratin in films mixed with meibum and compare these with the physicochemical characteristics of meibum obtained from sufferers of blepharitis. 

In conclusion, these data indicate that SPs and keratin likely interact with the TFLL. SPs are likely to have a stabilizing effect on the lipid layer and decrease the surface tension, while keratins are likely to destabilize the lipid layer. 

Acknowledgments

Supported by a University of Western Sydney Higher Degree Research Scholarship (CKP). 

Disclosure: C.K. Palaniappan, None; B.S. Schütt, Alcon (F), Allergan (F); L. Bräuer, None; M. Schicht, None; T.J. Millar, Alcon (F), Allergan (F)